arX

iv:1

204.

0907

v1 [

astr

o-ph

.GA

] 4

Apr

201

2

Mon. Not. R. Astron. Soc. 000, 1–15 (2011) Printed 5 April 2012 (MN LATEX style file v2.2)

High CO depletion in southern infrared-dark clouds

F. Fontani1⋆, A. Giannetti2, M.T. Beltran1, R. Dodson3, M. Rioja3, J. Brand2

P. Caselli4 and R. Cesaroni1

1 INAF-Osservatorio Astrofisico di Arcetri, L.go E. Fermi 5, Firenze, I-50125, Italy2 INAF-Istituto di Radioastronomia, via P. Gobetti 101, Bologna, I-40129, Italy3 International Centre for Radio Astronomy Research, University of Western Australia, Perth, Australia4 School of Physics and Astronomy, University of Leeds, Leeds, LS2 9JT , UK

Accepted date. Received date; in original form date

ABSTRACTInfrared-dark high-mass clumps are among the most promising objects to study theinitial conditions of the formation process of high-mass stars and rich stellar clus-ters. In this work, we have observed the (3–2) rotational transition of C18O withthe APEX telescope, and the (1,1) and (2,2) inversion transitions of NH3 with theAustralia Telescope Compact Array in 21 infrared-dark clouds already mapped in the1.2 mm continuum, with the aim of measuring basic chemical and physical parameterssuch as the CO depletion factor (fD), the gas kinetic temperature and the gas mass.In particular, the C18O (3–2) line allows us to derive fD in gas at densities higher(and hence potentially more depleted) than that traced by the (1–0) and (2–1) lines,typically used in previous works. We have detected NH3 and C18O in all targets. Theclumps have a median mass of ∼ 244 M⊙, are gravitationally bound, have an aver-age kinetic temperature of 17 K and possess mass, H2 column and surface densitiesconsistent with being potentially the birthplace of high-mass stars. We have measuredfD in between 5 and 78, with a mean value of 32 and a median of 29. These valuesare, to our knowledge, larger than the typical CO depletion factors measured towardsinfrared-dark clouds and high-mass dense cores, and are comparable to or larger thanthe values measured in low-mass pre–stellar cores close to the onset of the gravitationalcollapse. This result suggests that the earliest phases of the high-mass star and stellarcluster formation process are characterised by fD larger than in low-mass pre–stellarcores. On the other hand, fD does not seem to be correlated to any other physicalparameter, except for a faint anti-correlation with the gas kinetic temperature. Thir-teen out of 21 clumps are undetected in the 24 µm Spitzer images, and have slightlylower kinetic temperatures, masses and H2 column densities with respect to the eightSpitzer-bright sources. This could indicate that the Spitzer-dark clumps are either lessevolved or are going to form less massive objects.

Key words: Molecular data – Stars: formation – radio lines: ISM – submillimetre:ISM – ISM: molecules

1 INTRODUCTION

An ever increasing number of observational evidences indi-cates that the earliest phases of massive star and stellar clus-ter formation occur within infrared dark clouds (IRDCs).These are dense molecular clouds seen as extinction fea-tures against the bright mid-infrared Galactic background(e.g. Simon et al 2006, Ragan et al. 2006, 2011, Rathborneet al. 2007, 2010, Butler & Tan 2009). IRDCs are charac-terised by very high gas column densities (1023−1025cm−3)

⋆ E-mail: fontani@arcetri.astro.it

and low temperatures (6 25 K), so that they are believedto be the place where most of the stars in our Galaxy arebeing formed.

From an observational point of view, the spatial dis-tribution of the IRDCs in the Galaxy follows that of themolecular galactic component (with a concentration in theso-called 5 kpc molecular ring), and they are strong emittersof both far-IR/millimetre continuum and rotational molecu-lar transitions, especially those characterised by a high crit-ical density. In many of them, observations of the densegas at sub-parsec linear scales revealed the presence of on-going star formation, both in isolated and clustered mode

2 F. Fontani et al.

(e.g. Beuther & Sridharan 2007, Zhang et al. 2009, Fontaniet al. 2009, Jimenez-Serra et al. 2010, Pillai et al. 2011), in-cluding clear signs of high-mass star formation like hot cores(e.g. Rathborne et al. 2008) and/or Ultracompact Hii re-gions (Battersby et al. 2010). This demonstrates that IRDCsare indeed the birthplace of stars and stellar clusters of allmasses. Therefore, the IRDCs in the earliest evolutionarystages are the best locations where to study the initial con-ditions of the star formation process and put constraints oncurrent theories.

Despite the identification of thousands of IRDCs, thenumber of studies devoted to unveiling their physical andchemical properties remains still limited. There are veryfew targeted studies, especially in the southern hemisphere(e.g. Vasyunina et al. 2009, Miettinen et al. 2011), wherethere are fewer groundbased facilities operating in the (sub-)millimetre and centimetre domain than there are in thenorth. In particular, very little is known about the chemistryof IRDCs: studies suggest chemical compositions similar tothose observed in low-mass pre–stellar cores (Vasyunina etal. 2009), including large abundances of deuterated species(Pillai et al. 2007; Pillai et al. 2011; Fontani et al. 2011).However, the amount of CO freeze-out, a key chemical pa-rameter for pre–stellar cores (see e.g. Bergin & Tafalla 2007for a review) remains controversial: some works indicate highlevels (factor of 5, Hernandez et al. 2011) of CO freeze-out,while others do not reveal significant CO depletion (e.g. Mi-ettinen et al. 2011).

In this work we present observations of the rotationaltransition (3–2) of the dense gas tracer C18O, performedwith the Atacama Pathfinder EXperiment (APEX) 12-mTelescope, and of the inversion transitions (1,1) and (2,2)of NH3 carried out with the Australia Telescope CompactArray (ATCA), towards 21 IRDCs with declination lowerthan −30◦ already mapped in the 1.2 mm continuum. TheC18O observations allow to compute the CO depletion fac-tor, while the ammonia inversion transitions can be used toderive the temperature in dense and cold gas (see e.g. Ho &Townes 1983). Observations of N2H

+ and N2D+ in some of

the target sources, useful to derive the amount of deuteratedfraction (by comparing the N2D

+ and N2H+ column densi-

ties), are also presented. In Sect. 2 we describe the criteriaapplied to select the targets. Sect. 3 gives an overview of theobservations. The results are presented in Sect. 4 and dis-cussed in Sect. 5. Conclusions and a summary of the mainfindings are given in Sect. 6.

2 TARGET SELECTION

We selected 21 IRDCs from the 95 massive millimetreclumps detected by Beltran et al. (2006) in the 1.2 mm con-tinuum and non-MSX emitters (neither diffuse nor point-like). The targets were chosen according to these criteria:(i) source declination δ 6 −30◦; (ii) clumps isolated or hav-ing the emission peak separated by more than the SIMBAhalf power beam width to that of MSX-emitter objects, tolimit confusion and select the most quiescent sources; (iii)clump masses above ∼ 35M⊙ to deal with possible massivestar formation. In this work we have recomputed the massesutilising the gas temperature derived from ammonia for eachclump, and our results confirm that all targets are high-mass

clumps (see Sect 4.6). The list of IRDCs is given in Table 1.The coordinates correspond to the peak of the 1.2 millime-tre continuum emission mapped by Beltran et al. (2006).In Table 1 we also give some basic information like the dis-tance to the Sun, the Galactocentric distance, and the (non-)detection in the Spitzer-MIPS 24 µm images. A comparisonbetween the 1.2 mm continuum maps and the Spitzer 24 µmimages of each target (except for 13039–6108c6, for whichthe MIPS images are not available) is shown in Fig. A1 ofAppendix A.

3 OBSERVATIONS

3.1 APEX

Single-point spectra of the C18O (3–2), N2H+ (3–2) and

N2D+ (4–3) lines towards the sources listed in Table 1 were

obtained with the APEX Telescope in service mode betweenthe 20th and the 28th of June, 2008. The observations wereperformed in the wobbler-switching mode with a 150′′ az-imuthal throw and a chopping rate of 0.5 Hz. The receiverused for all lines was SHFI/APEX-2. The backend provideda total bandwidth of 1000 MHz. Details about the observedlines and the observational parameters are given in Table 2.The telescope pointing and focusing were checked regularlyby continuum scans on planets and the corrections were ap-plied on-line. Calibration was done by the chopper-wheeltechnique. Spectra were obtained in antenna temperatureunits (corrected for atmospheric attenuation), T ∗

A, and thenconverted to main beam temperature units through the re-lation TMB = Feff

BeffT ∗A (Feff and Beff are given in Col. 5 of

Table 2). The velocities adopted to centre the backends aregiven in Col. 6 of Table 1. For most sources we knew the ra-dial Local Standard of Rest velocities (VLSR) from previousCS observations. The spectra of 16435–4515c3 (for which wedid not have the CS data) were centred at VLSR = 0; in anycase the total bandwidth of the backend (∼ 1000 km s−1)was larger than the velocity gradient across the Galaxy.

3.2 ATCA

We observed the inversion transitions (J,K) = (1,1) and(2,2) of ammonia at 23694.5 MHz and 23722.6 MHz (K-bandat ∼ 1.2 cm), respectively, with the ATCA1 towards all tar-gets in Table 1. The observations were performed betweenthe 4th and the 8th of March, 2011, for a total telescope timeof 48 hours. We used the configuration 750D, which pro-vides baselines between 31 and 4469 m. The primary beamwas ∼ 2.5′ at the line frequencies. The flux density scalewas established by observing the standard primary calibra-tor 1934–638, and the uncertainty is expected to be of theorder of ∼ 10%. Gain calibration was ensured by frequentobservations of nearby compact quasars. The quasar 0537–441 was used for passband calibration. Pointing correctionswere derived from nearby quasars and applied online. Atmo-spheric conditions were generally good (weather path noise≃ 400 µm or better).

1 The Australia Telescope Compact Array is part of the AustraliaTelescope which is funded by the Commonwealth of Australia foroperation as a National Facility managed by CSIRO.

CO depletion in infrared dark clouds 3

Table 1. Source list and detection summary.

Source namea R.A. (J2000) Dec. (J2000) l b VLSRb dc DGC MIPS 24 µmd

h m s o ′ ′′ o o km s−1 kpc kpc

08477–4359c1 08:49:35.13 –44:11:59 264.69 –0.07 8.5 1.8 8.9 Y13039–6108c6 13:07:14.80 –61:22:55 305.18 1.14 –26.2 2.4 7.4 –13560–6133c2 13:59:33.04 –61:49:13 311.23 –0.35 –58.4 5.6 6.4 N14183–6050c3 14:22:21.54 –61:06:42 314.03 –0.52 –42.6 3.4 6.6 N15038–5828c1 15:07:32.52 –58:40:33 320.19 –0.77 –67.1 5.0 5.7 N15278–5620c2 15:31:44.17 –56:32:08 324.07 –0.73 –49.4 3.4 6.1 N15470–5419c1 15:51:28.24 –54:31:42 327.51 –0.83 –61.7 4.1 5.5 Ye

15470–5419c3 15:51:01.62 –54:26:46 327.51 –0.72 –61.7 4.1 5.5 Y15557–5215c2 15:59:36.20 –52:22:58 329.81 0.03 –67.6 4.4 5.2 Y15557–5215c3 15:59:39.70 –52:25:14 329.80 –0.00 –67.6 4.4 5.2 Nf

16061–5048c1 16:10:06.61 –50:50:29 332.06 0.08 –51.8 3.6 5.6 Y16061–5048c4 16:10:06.61 –50:57:09 331.98 0.00 –51.8 3.6 5.6 Nf

16093–5128c2 16:12:55.46 –51:43:22 331.77 –0.86 –97.3 6.1 4.3 N16093–5128c8 16:12:49.63 –51:36:34 331.84 –0.77 –97.3 6.1 4.3 Ng

16164–4929c3 16:20:24.51 –49:35:34 334.11 –0.17 –33.8 2.6 6.3 Ng

16254–4844c1 16:29:00.89 –48:50:31 335.63 –0.65 –45.5 3.4 5.6 Y16435–4515c3 16:47:33.13 –45:22:51 340.31 –0.71 0.0 –h 4.9 N16482–4443c2 16:51:44.59 –44:46:50 341.24 –0.90 –43.3 3.7 5.1 Nf

16573–4214c2 17:00:33.38 –42:25:18 344.08 –0.67 –23.7 2.6 6.0 Y17040–3959c1 17:07:58.78 –40:02:24 346.82 –0.35 –0.4 16.5i 8.5 N?l

17195–3811c2 17:23:00.30 –38:14:58 349.97 –1.68 –26.4 3.6 5.0 Y

a derived from the IRAS source name and the ’clump’ number assigned by Beltran et al (2006);b from the CS (2–1) or (3–2) transitions (Fontani et al. 2005);c kinematic distance from the Sun (Beltran et al 2006). In case of distance ambiguity, the near distance is adopted;d detection (Y) or non-detection (N) of the clump in the Spitzer-MIPS 24µm band;e point-like infrared source not at clump centre, probably associated with it;f point-like infrared source at the edge of the 3 σ rms contour level of the millimetre continuum, probably not associated with it;g diffuse infrared emission inside the 3 σ rms contour level of the millimetre continuum, probably associated with a nearby infrared object;h the kinematic distance recomputed in Sect. 4.6 from the C18O (3–2) line peak velocity is 3.1 kpc;i near distance smaller than 100 pc, so that the far distance is adopted;l the millimetre map of the clump is incomplete because at the edge of the SIMBA field of view, thus one cannot conclusively assert if theclump is associated with an infrared source (see Fig. A1).

Table 2. Transitions observed with APEX and observational parameters.

Transition νa HPBW ∆V b Beff/Feff Tsys

GHz ′′ km s−1 K

C18O (3–2) 329.331 18.9 ∼ 0.22 0.74/0.95 200 – 300c

N2D+ (4–3) 308.422d 20.2 ∼ 0.23 0.74/0.95 160 – 300N2H+ (3–2) 279.512e 22.3 ∼ 0.26 0.74/0.95 130 – 160

a line rest frequency;b spectral resolution;c for 13039–6108c6, 13560–6133c2 and 14183–6050c3, Tsys ∼ 600 K;d from the Cologne Database for Molecular Spectroscopy (CDMS), and it refers to the F1F = 5 5 → 4 4hyperfine component;e from Pagani et al. (2009), and it refers to the F1F = 4 5 → 3 4 hyperfine component, which has a relativeintensity of 17.46%.

The total time was broken up into series of 3-/5-minutesnapshots in order to improve the coverage of the uv-planefor each target. As a consequence of this observing strategy,the integration time on source is variable, and generally inbetween ∼ 30 mins and ∼ 1 hour. The CABB2 correlatorprovided two ”zoom” bands of 64 MHz each, with a spec-

2 http://www.narrabri.atnf.csiro.au/observing/CABB.html

tral resolution in each zoom of 32 kHz (∼0.4 km s−1 at thefrequencies of the lines). The ammonia lines were observedin one zoom band. The data were edited and calibrated fol-lowing standard tasks and procedures of the MIRIAD soft-ware package. After the editing and calibration in MIRIAD,the data were imported in AIPS. Imaging and deconvolu-tion were performed using the ’imagr’ task, applying natu-ral weighting to the visibilities. The ammonia emission wasdetected only on the short baselines, thus we discarded all

4 F. Fontani et al.

baselines > 30 kλ. In order to obtain images with the sameangular resolution, and comparable to the APEX beam, wereconstructed the images with a circular beam of 20′′ ofdiameter for all sources, except for 17195-3811c2 and 17040-3959c1, that have a poorer UV-coverage, and for which weget a beam of roughly 20′′ × 40′′. Moreover, 16254-4844c1and 16573-4214c2 were observed only once, making the cleanimpossible and thus forcing us to use the dirty image decon-volved with the dirty beam to determine the spectrum.

In this work the ATCA data will be used only to derivethe gas temperature from the NH3 (1,1) and (2,2) spectraat the dust emission peak (in Sect. 4.5). A complete presen-tation and analysis of the ATCA observations will be givenin a forthcoming paper (Giannetti et al., in prep).

4 RESULTS

4.1 Detection summary

The detection summary is given in Table 3. C18O (3–2) wasobserved and detected in all sources. Twelve targets werealso observed and detected in N2H

+ (3–2). Ten clumps wereobserved in N2D

+ (4–3), and only clump 15557–5215c2 wasmarginally detected. This is, to our knowledge, the first de-tection of this line towards an IRDC. All targets were ob-served and detected in the ammonia inversion transitions,with the exception of 14183–6050c3 which was undetectedin the (2,2) line.

4.2 Line profiles and parameters

All spectra of the C18O and N2H+ lines are shown in Ap-

pendix B. Fig. B1 shows the N2H+ (3–2) and C18O (3–2)

transitions in the twelve clumps observed and detected inboth lines. Fig. B2 shows the nine sources observed and de-tected in C18O (3–2) only.

Most of the C18O lines are well fitted by single Gaus-sians. The results of these fits are reported in Table 4.Slightly asymmetric profiles deviating from the Gaussianshape can be noted in 15470–5419c1, 16061–5048c1, 16482–4443c2, 16573–4214c2 and 17040–3959c1. This could bedue either to the superposition of multiple blended ve-locity components, or to high optical depth effects. Non-Gaussian wings in one (or both) side(s) of the line indicat-ing presence of outflows are noticeable towards two sources,08477–4359c1 and 17195–3811c2 (see Fig. 1). Four sourcesshow multiple lines well-separated in velocity, 14183–6050c3,16093–5128c2, 16093–5128c8, 16435–4515c3. Because wehad the previous CS observations (Fontani et al. 2005) asreference, in Table 4 we have identified the line associatedwith the star-forming region of our interest. The other com-ponents are almost certainly arising from clouds along theline of sight but not associated with the star-forming re-gion, given the large separation in velocity (from ∼ 20 to∼ 40 km s−1, see Figs. B1 and B2).

The N2H+ (3–2) lines roughly peak at the same position

as the C18O (3–2) ones and have comparable line widths.Because of the hyperfine structure present in the transition,it is not straightforward to derive conclusions on blendedmultiple velocity components. However, we note hints ofpossible secondary velocity components in all the lines that

Figure 1. C18O (3–2) spectra of 08477–4359c1 and 17195–3811c2. The Gaussian fits are superimposed on the spectraand highlight significant emission in non-Gaussian wings.

Table 4. C18O line parameters derived from Gaussian fits. For sourceswith more than one velocity component, the line labelled as ’–a’ is theone associated with the IRDC (based on previous CS observations,Fontani et al. 2005).

source∫

TMBdv vLSR FWHM Tpeak

K km s−1 km s−1 km s−1 K

08477–4359c1 6.10(0.09) 9.53(0.02) 2.44(0.04) 2.313039–6108c6 1.4(0.1) -26.04(0.06) 1.8(0.1) 0.7613560–6133c2 1.6(0.15) -57.47(0.08) 1.9(0.2) 0.814183–6050c3–a 0.9(0.1) -45.52(0.08) 1.5(0.2) 0.5614183–6050c3–b 0.57(0.06) -26.70(0.02) 0.48(0.05) 1.115038–5828c1 0.9(0.1) -69.3(0.3) 5.2(0.7) 0.1615278–5620c2 2.5(0.1) -49.42(0.04) 2.0(0.1) 1.1815470–5419c1 2.3(0.1) -58.74(0.07) 2.45(0.2) 0.8815470–5419c3 2.6(0.1) -61.38(0.06) 2.9(0.1) 0.8415557–5215c2 4.2(0.1) -67.57(0.03) 2.28(0.08) 1.715557–5215c3 1.7(0.1) -68.47(0.08) 2.4(0.2) 0.6716061–5048c1 7.61(0.02) -67.28(0.03) 3.58(0.06) 2.016061–5048c4 3.2(0.1) -51.99(0.03) 1.94(0.07) 1.616093–5128c2–a 3.6(0.1) -96.19(0.05) 3.4(0.1) 1.016093–5128c2–b 1.97(0.07) -63.46(0.03) 1.57(0.07) 1.216093–5128c8–a 0.96(0.06) -96.85(0.08) 2.3(0.2) 0.3916093–5128c8–b 0.19(0.04) -68.9(0.1) 1.0(0.2) 0.1816164–4929c3 2.95(0.07) -33.80(0.02) 1.76(0.05) 1.616254–4844c1 2.72(0.09) -44.98(0.03) 2.20(0.09) 1.216435–4515c3–a 1.7(0.1) -35.57(0.04) 1.6(0.1) 1.016435–4515c3–b 0.88(0.09) -52.09(0.06) 1.3(0.2) 0.6216435–4515c3–c 0.5(0.1) -24.4(0.1) 1.5(0.4) 0.3116482–4443c2 4.6(0.1) -43.35(0.05) 3.3(0.1) 1.316573–4214c2 8.1(0.2) -27.29(0.03) 2.76(0.06) 2.817040–3959c1 6.45(0.08) -15.87(0.02) 2.40(0.03) 2.517195–3811c2 2.54(0.07) -26.41(0.02) 1.47(0.06) 1.6

show similar features in the C18O (3–2) line (15470–5419c1,16061–5048c1, 16573–4214c2) except for 16482–4443c2.

In Table 5 we give the N2H+ (3–2) line parameters: in

Cols. 3 – 7 we list integrated intensity (∫

TMBdv), peak ve-locity (VLSR), FWHM, opacity (τ ), and excitation tempera-ture (Tex) of the N2H

+ (3–2) line, respectively. Because therotational transitions of N2H

+ possess hyperfine structure,we fitted the lines using the METHOD HFS of the CLASS

CO depletion in infrared dark clouds 5

Table 3. Detection summary: Y = detected, N = not detected, – = not observed;

Source APEX ATCA (NH3)C18O (3–2) N2H+ (3–2) N2D+ (4–3) NH3 (1,1) / (2,2)

08477–4359c1 Y Y N Y / Y13039–6108c6 Y Y N Y / Y13560–6133c2 Y – – Y / Y14183–6050c3 Y – – Y / N15038–5828c1 Y – N Y / Y15278–5620c2 Y – N Y / Y15470–5419c1 Y Y N Y / Y15470–5419c3 Y Y N Y / Y15557–5215c2 Y Y Ya Y / Y15557–5215c3 Y Y N Y / Y16061–5048c1 Y Y N Y / Y16061–5048c4 Y Y N Y / Y16093–5128c2 Y Y – Y / Y16093–5128c8 Y – – Y / Y16164–4929c3 Y – – Y / Y16254–4844c1 Y – – Y / Y16435–4515c3 Y Y – Y / Y16482–4443c2 Y Y – Y / Y16573–4214c2 Y Y – Y / Y17040–3959c1 Y – – Y / Y17195–3811c2 Y – – Y / Y

a marginal detection (see Sect. 4.4).

Table 5. N2H+ line parameters and column densities. For optically thin transitions without well-constrained opacity(τ = 0.1), to compute N(N2H+) we adopted a Tex= 6 K (see text). For sources with more than one velocity component,the line labelled as ’–a’ is the one having the peak velocity consistent with that of the clump of our interest, as in Table 4.

source vel. range∫

TMBdv vLSR FWHM τ Tex N(N2H+)

km s−1 K km s−1 km s−1 km s−1 K (×1013)cm−2

08477–4359c1 5.4,14.8 6.02(0.07) 10.53(0.02) 2.87(0.05) 0.1 6 2.93(0.03)13039–6108c6 –30.0,–23.6 0.66(0.08) –26.2(0.1) 1.6(0.4) 0.1 6 0.32(0.04)15470–5419c1 –66.1,–54.8 5.2(0.1) –58.82(0.03) 2.4(0.2) 12(2) 4.8(0.3) 54(1)15470–5419c3 –67.1,–56.8 9.4(0.1) –61.38(0.03) 3.14(0.02) 0.1 6 4.6(0.05)15557–5215c2 –74.1,–62.6 17.2(0.2) –66.77(0.02) 2.01(0.05) 12.7(0.8) 8.0(0.5) 27(2)15557–5215c3 –71.8,–66.2 1.12(0.07) –68.69(0.06) 1.61(0.09) 1.4(0.2) 4.6(0.2) 1.8(0.1)16061–5048c1 –72.6,–59.5 25.3(0.2) –66.20(0.02) 3.14(0.09) 9.1(0.8) 9(1) 29(3)

16061–5048c4 –56.3,–48.3 2.13(0.08) –51.98(0.05) 1.29(0.09) 7(1) 4.5(0.3) 18(4)16093–5128c2–a –102.1,–92.1 1.28(0.06) –97.20(0.08) 3.1(0.2) 0.1 6 0.62(0.03)16093–5128c2–b –68.0,–60.2 0.63(0.06) –63.21(0.08) 1.7(0.3) 0.3(1.5) 6 0.38(0.04)16435–4515c3–a –53.0,–49.8 0.08(0.02) –51.1(0.3) 1.6(0.4) 0.1 6 0.04(0.01)16435–4515c3–b –37.1,–32.5 0.23(0.02) –35.2(0.1) 1.5(0.3) 1(0.2) 3.5(0.7) 2.7(0.2)16482–4443c2 –47.0,–38.6 3.68(0.07) –42.57(0.03) 3.02(0.09) 0.1 6 1.79(0.03)16573–4214c2 –32.5,–22.2 9.52(0.08) –27.06(0.01) 3.79(0.02) 0.1 6 4.64(0.04)

package3. This method fits all the hyperfine components si-multaneously assuming that they have the same excitationtemperature and width, that the opacity has a Gaussian de-pendence on frequency and fixing the separation of the com-ponents to the laboratory value. The line parameters listedin Table 5 have been derived from this method, except theintegrated intensity that has been computed by simple in-tegration over the velocity range given in Col. 2. The lineopacity is deduced from the intensity ratio of the different

3 The CLASS program is part of the GILDAS software, devel-oped at the IRAM and the Observatoire de Grenoble, and is avail-able at http://www.iram.fr/IRAMFR/GILDAS

hyperfine components, most of which are blended due to thefact that the line widths are larger (∼ 1 − 3 km s−1) thanthe separation in velocity of the components. However, thefit residuals are generally low indicating that the procedureprovides good results despite the blending.

4.3 Column densities of N2H+

The N2H+ column densities were derived following the

method outlined in the Appendix A of Caselli et al. (2002a).Specifically, we adopted equations (A-1) and (A-4) for opti-cally thick and optically thin lines, respectively. The methodassumes a constant excitation temperature, Tex. An estimate

6 F. Fontani et al.

of Tex can be derived from the fitting procedure to the hyper-fine structure 4, but not for optically thin lines or lines withopacity not well-constrained. For these, we have assumedthe average Tex (∼ 6 K; Table 5) derived for the sourceswith well-constrained opacity.

For all sources but 16573–4214c2 and 16482–4443c2,the clump diameter derived from the 1.2 mm continuumis comparable to or larger than the beam size (see Table 2in Beltran et al. 2006, see also Table 8), so that we assumedunity filling factor. This factor was applied also to the un-resolved source 16482–4443c2. We stress that this is an ap-proximation, because the sources could be clumpy. However,due to the lack of observations at higher-angular resolution,neither in the lines observed in this work nor in other high-density gas tracers, the size of the effective emitting regioncannot be determined. For 16573–4214c2, for which θs ∼ 7′′,we applied a correction factor 1/ην = (θ2s +HPBW 2)/θ2s ∼10.

The N2H+ total column densities are given in Col. 8 of

Table 5, and span a range from 1012 to 1014 cm−2. Thesevalues are in good agreement with those derived from thesame line in other IRDCs (Ragan et al. 2006, Miettinenet al. 2011), as well as in other massive starless and star-forming cores (Fontani et al. 2011, Chen et al. 2011).

4.4 Upper limits on the column density of N2D+

and on N2D+/N2H

+ column density ratio

The N2D+ (4–3) transition was marginally detected at ∼

3.2 σ rms only towards 15557–5215c2. This represents, toour knowledge, the first detection of this line in an IRDC.The spectrum is shown in Fig. 2. For this source we havecomputed the N2D

+ column density fitting the line with aGaussian (the results are given in Table 6), and then follow-ing the approach of Appendix A in Caselli et al. (2002a) toderive the N2D

+ column density for the optically thin case.This line is also a blend of hyperfine components, thereforethe line width derived from the Gaussian fit is an upperlimit to the intrinsic value. However, because of the poorsignal-to-noise ratio of the spectrum, METHOD HFS didnot provide reliable results. As Tex, we used that derivedfrom N2H

+ (3–2) for this source (see Table 5). We findN(N2D

+)∼ 8×1011cm−2 and the corresponding deuteratedfraction (column density ratio N(N2D

+)/N(N2H+) =Dfrac)

is ∼ 0.003 (Table 6). This value is consistent with studiesperformed in both IRDCs (e.g. Miettinen et al. 2011, Chen etal. 2011) and massive star-forming clumps containing moreevolved objects (e.g. Fontani et al. 2006, 2011).

For all the other undetected sources, we have derivedupper limits on the deuterated fraction in between 0.003 and0.5. These upper limits were computed from the integratedintensity upper limits assuming the lines to be Gaussianfrom the formula

∫

TMBdv = ∆V

2√

ln2/πT peakMB . We adopted the

3σ rms level in the spectrum as peak temperature T peakMB , and

the line width measured in the detected source 15557–5215c2(Table 2) as ∆V . The Dfrac upper limits are comparable tothe values commonly measured in massive clumps (Fontani

4 See the CLASS user manual for the deriva-tion of Tex from the output parameters:http://iram.fr/IRAMFR/GILDAS/doc/html/class-html/class.html/

Figure 2. N2D+ (4–3) spectrum of the only source detectedin this transition, 15557–5215c2.

et al. 2006, Chen et al. 2011, Miettinen et al. 2011), so thatour sensitivity does not allow to conclude if our targets havedeuterated fraction similar to or lower than the other clumpsin the literature. We believe that the very low detection ratein the N2D

+ (4–3) line is likely due to the high densities re-quired to excite this transition: having a critical density of∼ 3×107 cm−3, this line is expected to come from very com-pact regions, thus suffering enormous beam dilution effects.

4.5 Rotation temperatures and NH3 columndensities at dust emission peak position

We have extracted spectra of the NH3 (1,1) and (2,2) tran-sitions from the ATCA channel maps towards the positionsgiven in Table 1. The spectra obtained this way were thenimported in CLASS, and fitted using METHOD NH3 to fitthe (1,1) lines, which takes into account the line hyperfinestructure, similarly to METHOD HFS (see Sect. 4.2). Thiswas also used for the (2,2) emission, even though the hyper-fine components were not always visible, to obtain an upperlimit for τ .

In Table 7 we list the rotation temperatures (Trot), ki-netic temperatures (Tk), and total column densities of am-monia (NNH3

) derived from the NH3 (2,2)/(1,1) line ra-tios, as well as the line peak velocities. We have derivedTrot and NNH3

from the output parameters given by the fit-ting method outlined above, and using the formulae givenin the Appendix of Busquet et al. (2009). The formulae,which result from the discussion in Ho & Townes (1983, Eq.(4)), have been derived assuming that the transitions be-tween the metastable inversion doublets are approximatedas a two-level system, and that the excitation temperatureand line width are the same for both NH3 (1, 1) and NH3

(2, 2). Note that the assumption of a two-level system isreasonable because transitions between the metastable in-version doublets are usually much faster than those of otherrotational states (Ho & Townes 1983). Tk was extrapolatedfrom Trot following the empirical approximation method de-scribed in Tafalla et al. (2004, see also Eq. (A.5) in Busquetet al. 2009). We find kinetic temperatures in between 13and 25 K, with both mean and median Tk of ∼ 17 K. Thesevalues are consistent, within the errors, with typical tem-peratures measured towards IRDCs (Pillai et al. 2006) fromammonia.

CO depletion in infrared dark clouds 7

Table 6. N2D+ (4–3) line parameters and N2D+ column density in 15557–5215c2

∫

TMBdv vLSR FWHM Tpeak N(N2D+) N(N2D+)/N(N2H+)

K km s−1 km s−1 km s−1 K (×1011)cm−2

0.14( 0.03) –68.1(0.2) 2.0(0.5) 0.07 8(2) 0.003(0.001)

Table 7. Rotation temperature, kinetic temperature, totalNH3 column densities and peak velocities derived from theNH3 inversion transitions observed with the ATCA. The er-rors on Trot and Tk are given in parentheses and are computedfrom the propagation of errors. The uncertainty on the columndensities includes the error on the flux calibration, and is es-timated to be of the order of 20 – 30%.

Source Trot Tk NNH3Vpeak

a

K K 1015 cm−2 km s−1

08477–4359c1 16(2) 19(3) 1.06 +11.18(0.02)13560–6133c2 15(9) 17(10) 4.61 –56.32(0.05)14183–6050c3b – – – –44.36(0.04)15038–5828c1 14(3) 15(4) 2.97 –68.72(0.04)15278–5620c2 17(2) 20(3) 3.52 –48.69(0.02)15470–5419c1–a 16(2) 18(3) 4.23 –60.72(0.03)15470–5419c1–b 13(9) 14(9) 3.20 –57.88(0.06)15470–5419c3 17(1) 19(2) 3.61 –61.02(0.01)15557–5215c2 19(2) 23(4) 4.62 –67.11(0.01)15557–5215c3 14(3) 15(3) 4.47 –68.53(0.01)16061–5048c1 20(3) 25(5) 4.44 –66.76(0.01)

16061–5048c4 12(2) 13(2) 5.28 –51.62(0.01)16093–5128c2 14(5) 16(5) 3.87 –97.09(0.05)16093–5128c8 14(6) 15(7) 1.09 –95.80(0.04)16164–4929c3 13(6) 14(6) 2.59 –32.80(0.01)16254–4844c1c 18(4) 22(6) 4.77 -38.45(0.05)16435–4515c3 11(4) 12(5) 1.84 –51.69(0.04)16482–4443c2 15(3) 16(4) 0.78 –42.35(0.03)16573–4214c2c 15(4) 17(5) 3.32 -25.09(0.04)17040–3959c1 16(3) 18(4) 6.81 –15.62(0.02)17195–3811c2 15(1) 17(2) 1.13 –25.51(0.01)

a derived from the (1,1) transition. The uncertainty (betweenparentheses) is that given by the fitting procedure;b undetected in the (2,2) transition (see Table 3);c observed only for one cycle with bad UV coverage. Thevalues listed have been derived from the spectra extractedfrom the dirty map deconvolved with the dirty beam.

4.6 H2 masses and other physical parametersfrom 1.2 mm continuum emission

Gas masses were calculated by Beltran et al. (2006) for allclumps from the 1.2 mm continuum integrated flux den-sity, assuming optically thin emission and a reasonable dusttemperature of 30 K for all clumps. This latter assumptionwas due to the fact that a temperature estimate for eachclump was lacking. In this work we can profit from the tem-peratures derived from ammonia (Col. 3 of Table 7) andrecompute the H2 masses assuming that the dust temper-ature equals the kinetic temperature. This method impliescoupling between gas and dust, which is a realistic assump-tion for gas densities as those of our clumps (> 104 cm−3).

The gas mass, M , has been derived using Eq. (1) inBeltran et al. (2006), and adopting the same assumptions,

except the dust temperature for which we have utilised Tk

derived from ammonia. From M , for each source we havecalculated the source averaged H2 volume and column den-sities, nH2

and Nt(H2), assuming the clumps to be sphericaland homogeneous. The H2 column densities, Np(H2), havebeen estimated also from the 1.2 mm continuum peak fluxusing the equations in Beuther et al. (2002) and adoptingthe same assumptions made to derive the mass (same β, dustmass opacity and gas-to-dust ratio). These represent aver-age values in the telescope beam of the 1.2 mm continuumobservations (∼ 24′′), and we will use these estimates to de-rive the CO depletion in Sect. 4.7, because the APEX C18Oobservations have a comparable angular resolution (∼ 19′′).Also, we have computed the source averaged surface den-sity, Σ(H2) = 4M/π(θsd)

2, where d and θs are the sourcedistance (see Table 1) and angular diameter (see Table 8),respectively. For 16435–4515c3, for which we did not have akinematic source distance, we computed it from the velocityof the C18O (3–2) line of the ’–a’ component (see Table 4)following the method explained in Fontani et al. (2005). Themethod, which assumes the rotation curve of Brand & Blitz(1993), is valid for distances from the Galactic centre be-tween 2 and 25 kpc, and provides two kinematic distances,3.1 and 12.6 kpc. As for the other sources with distanceambiguity, we adopted the ‘near’ value, i.e. 3.1 kpc.

The results are listed in Table 8, where we give M ,nH2

, N(H2) (both Nt(H2) and Np(H2)) and Σ(H2). For com-pleteness, in Cols. 2 and 3 of Table 8 we list the full widthhalf maximum diameters, θs (deconvolved with the telescopebeam of 24′′) adopted to derive nH2

, Nt(H2) and Σ(H2),as well as the integrated flux densities used to calculatethe masses, Sν . Both parameters are taken from Beltran etal. (2006). The mean mass turns out to be 439 M⊙, and themedian is 244 M⊙. As expected, these values are systemati-cally higher than those derived by Beltran et al. (2006) giventhat the kinetic temperatures from ammonia (see Table 7)are systematically lower than the representative value (30 K)assumed by Beltran et al. (2006). The clump-averaged col-umn densities are of the order of 1022 − 1023 cm−2, with amean value of 1.6 × 1023 cm−2 and a median of 8.5 × 1022

cm−2. The volumn densities are of the order of 104 − 105

cm−3, with a mean value of 3.1 × 105 cm−3 and a medianof 3.9× 104 cm−3. Finally, the mean surface density is 0.37g cm−2, and its median value is 0.19 g cm−2. The differencebetween mean and median values for all parameters is dueto the clump 16573–4214c2, which has an angular diameter(∼ 7′′) more than twice smaller than any other source of thesample, despite the comparable mass.

The clump-averaged column densities Nt(H2) are gen-erally smaller than the theoretical threshold given byKrumholz & McKee (2008) to avoid cloud fragmentation,which is ∼ 1 g cm−2, corresponding to ∼ 3 × 1023 cm−2.Also, the distance independent parameter Σ(H2) is smaller,on average, than the theoretical values predicted for clumps

8 F. Fontani et al.

that are going to form high-mass stars or super star clus-ters, which are expected to be larger than ∼ 0.7 g cm−2

(Chakrabarti & McKee 2005, Krumholz & McKee 2008).Rather, the values measured in this work are consistentto those predicted for clumps that are going to formintermediate-mass stars and stellar clusters (Chakrabarti &McKee 2005). We stress, however, that Σ(H2) and Nt(H2)represent average values across the whole clumps, whichcould in reality be fragmented in smaller and denser cores.Therefore, our Σ(H2) and Nt(H2) are to be considered aslower limits for the individual embedded cores. On the otherhand, the column densities calculated from the 1.2 mmcontinuum peak flux, Np(H2), are closer to the theoreticalthreshold proposed by Krumholz & McKee (2008), indicat-ing that the central regions of the clumps could more easilyform massive stars. Moreover, we have compared the clumpmasses to the threshold proposed by Kauffmann & Pil-lai (2010) based on an empirical mass-radius relation whichpredicts that a cloud must exceed the relation M > Mthr =870M⊙(R/pc)1.33 to form massive stars. The mass thresh-olds Mthr are listed in Col. 9 of Table 8. As R, we have usedhalf of the diameters listed in the same Table. We can seethat our clump masses are all above the predicted threshold,with four exceptions: 13039–6108c6, 15557–5215c3, 16164–4929c3 and 16482–4443c2, for which, in any case, the mea-sured mass and the corresponding threshold value are sim-ilar, and for 16482–4443c2 the mass threshold is even anupper limit. Based on these results, we claim that the tar-gets have the potential to form massive stars.

4.7 C18O column densities and CO depletionfactors

The C18O column density has been derived from line in-tensities assuming optically thin lines and LTE conditions.Under these assumptions one can demonstrate that the totalcolumn density of C18O, NC18O, is related to the integratedintensity of a rotational transition J → J − 1 according to(e.g. Pillai et al. 2007):

NC18O =NJ

gJQ(Tex)e

(

EJ

kTex

)

=

=3h

8π3

1

Sµ2

I(C18O)

Jν(Tex)− Jν(TBG)

Q(Tex)e

(

EJ

kTex

)

ehν/kTex − 1

1

ην(1)

where: I(C18O) is the integrated intensity of the line; NJ

is the column density of the upper level; gJ , EJ and S arestatistical weight (=2J + 1), energy of the upper level andline strength, respectively; Q(Tex) is the partition functionat the temperature Tex; ν the line rest frequency; Jν(Tex)and Jν(TBG) the equivalent Rayleigh-Jeans temperatures atfrequency ν computed for the excitation and backgroundtemperature (TBG ∼ 2.7 K), respectively; µ the molecule’sdipole moment (0.1102 Debye for C18O); ην the beam fillingfactor. The total integrated intensity I(C18O) was derivedfrom the integral over all the channels with emission insteadof the area given by the Gaussian fits to take into accountalso non-Gaussian features.

As for the derivation of the N2H+ column density, we

assumed a unity filling factor for all sources, so thatNC18O isa beam-averaged value. As excitation temperature, we haveused the kinetic temperatures listed in Col. 3 of Table 7

(i.e. we are assuming LTE conditions for C18O (3–2)). Theresulting column densities are listed in Table 9. To check ifthe assumption of optically thin lines is reasonable, we haveestimated the line optical depths from the solution of theline radiative transfer equation:

τ = −ln

(

1−TMB

Jν(Tex)− Jν(TBG)

)

and found values smaller than ≃ 0.6, consistent with ourassumption of optically thin lines.

The CO depletion factor, fD, is defined as the ratiobetween the ’expected’ abundance of CO relative to H2

(XEC18O) and the ’observed’ value:

fD =XE

C18O

XOC18O

. (2)

XOC18O is the ratio between the observed C18O column den-

sity and the observed H2 column density. For this latter,we have used the value derived from the 1.2 mm continuumpeak flux, Np(H2) (Col. 7 of Table 8), which is averagedover a beam comparable to that of the C18O observations(24′′ against 19′′).

To compute XEC18O, we have taken into account the

variation of atomic carbon and oxygen abundances withdistance to the Galactic Center following the same proce-dure adopted in Fontani et al. (2006; see also Miettinen etal. 2011). Assuming the canonical abundance of ∼ 8.5×10−5

for the abundance of the main CO isotopologue in the neigh-bourhood of the solar system (Frerking et al. 1982, see alsoLanger et al. 1989 and Pineda et al. 2008), we have computedthe expected CO abundance at the Galactocentric distance(DGC) of each source according to the relationship:

XEC16O = 8.5× 10−5exp (1.105 − 0.13DGC(kpc)) , (3)

which has been derived according to the abundance gradi-ents in the Galactic Disk for 12C/H and 16O/H listed inTable 1 of Wilson & Matteucci (1992), and assuming thatthe Sun has a distance of 8.5 kpc from the Galactic Centre.Then, following Wilson & Rood (1994), we have assumedthat the Oxygen isotope ratio 16O/18O depends on DGC ac-cording to the relationship 16O/18O=58.8×DGC(kpc)+37.1,which gives:

XEC18O =

XEC16O

(58.8DGC + 37.1). (4)

The results are listed in Table 9: Cols. 2, 3 and 4list the C18O (3–2) integrated line intensity (I(C18O)), theC18O column density (N(C18O)) and the observed C18Oabundance (XO

C18O); Cols. 5 and 6 give the expected C18Oabundance (XE

C18O) calculated at the Galactocentric dis-

tance of the source (Table 1) and the CO depletion factor(fD = XE

C18O/XOC18O), respectively.

The mean fD is ∼ 32 and the median value is 29. Thesevalues are remarkably higher than those obtained in otherIRDCs (Pillai et al. 2007, Miettinen et al. 2011, Hernandezet al. 2011), and are among the highest obtained both in low-mass starless cores (Crapsi et al. 2005, Tafalla et al. 2006)and in massive clumps from observations with low-angularresolution (Fontani et al. 2006, Chen et al. 2011).

Such a difference could be explained by the fact thatthe transition used in this work to derive fD has a criti-cal density of ∼ 5 × 104 cm−3, comparable to that of the

CO depletion in infrared dark clouds 9

Table 8. Parameters derived from the 1.2 mm continuum emission: 1.2 mm continuum flux density (Sν) integrated overthe 3σ rms contour level, clump angular diameter (θs), gas mass (M), H2 volume (nH2

), column (N(H2)) and mass surfacedensity (Σ(H2)). The last column lists the mass threshold, Mthr, for a cloud that can form massive stars based on theempirical mass-radius relation proposed by Kauffmann & Pillai (2010, see text).

source Sνa θsb M nH2

Nt(H2)c Np(H2)d Σ(H2) Mthr

(Jy) (′′) (M⊙) (×104 cm−3) (×1022 cm−2) (×1023 cm−2) (g cm−2) (M⊙)

08477–4359c1 1.8 35.6 86.73 11.5 11.0 1.42 0.24 7313039–6108c6 1.04 40.3 101.5 3.90 5.64 0.68 0.12 12713560–6133c2 0.77 23.8 426.8 6.27 12.5 1.09 0.27 19414183–6050c3 1.06 39.6 207.7 2.96 5.95 0.68 0.13 19615038–5828c1 0.49 24.7 250.6 4.62 8.53 0.85 0.19 17515278–5620c2 1.5 40.1 243.7 3.34 6.81 0.82 0.15 20015470–5419c1 1.16 24.2 310.2 11.0 16.4 1.37 0.36 13115470–5419c3 3.06 54.1 743.4 2.37 7.84 1.11 0.17 38215557–5215c2 2.89 41.3 633.4 3.67 9.96 1.55 0.22 29315557–5215c3 0.51 35.8 194.3 1.73 4.07 0.49 0.09 24216061–5048c1 2.1 28.1 284.3 9.54 14.4 1.66 0.31 134

16061–5048c4 1.57 62.8 504.2 1.52 5.12 1.22 0.11 39116093–5128c2 0.67 39.6 481.6 1.19 4.29 0.92 0.09 42716093–5128c8 0.86 40.1 642 1.52 5.57 0.68 0.12 43416164–4929c3 0.35 36.2 54.79 2.28 3.21 1.50 0.07 12216254–4844c1 1.15 20.3 164.2 17.4 17.9 1.43 0.39 8116435–4515c3 0.51 17.7 147 30.9 25.3 1.20 0.55 6116482–4443c2 0.24 ≪ 24e 59.08 ≫ 3.81e ≫ 4.63e 0.66 ≫ 0.10e ≪ 101e

16573–4214c2 0.93 7.29 108.3 553 157 1.89 3.4 1417040–3959c1 0.81 16.6 3428 5.80 23.7 1.60 0.52 50517195–3811c2 2.61 44.3 597.9 5.12 12.2 1.77 0.27 246

a from Beltran et al. (2006);b full width half maximum angular diameters deconvolved with the telescope beam, from Beltran et al. (2006);c computed from the total clump mass (see text);d computed from the 1.2 mm continuum peak flux (see text);e unresolved in the SIMBA map. Therefore, nH2

, Nt(H2) and Σ(H2) are lower limits, and Mthr is an upper limit.

central region where the CO freeze-out is expected to beimportant. In fact, the CO freeze-out timescale, which de-pends on the H2 volume density, becomes shorter than thefree-fall timescale at densities larger than a few 104 cm−3

(see e.g. Bergin & Tafalla 2007, Caselli 2011). On the con-trary, in most of the studies performed so far fD was derivedfrom the CO (1–0) or (2–1) lines, which trace lower-densitygas where CO is significantly non-depleted, resulting in val-ues of fD smaller than or comparable to ∼ 10 (e.g. Crapsiet al. 2005, Tafalla et al. 2006, Pillai et al. 2007). However,the discussion of this result requires three main comments.First, the values of the ’canonical’ CO abundance measuredby other authors in different objects vary by a factor of 2(see e.g. Lacy et al. 1994; Alves et al. 1999), but becausethe value assumed by us was derived from the Taurus starforming regions (Frerking et al. 1982) and used in previousestimates (Crapsi et al. 2005; Emprechtinger et al. 2009), itis likely the most appropriate to make comparison with low-mass dense cores. Second, the angular resolution of our ob-servations allows us to derive only average values of fD overangular regions much larger than the typical fragmentationscales (few arcseconds at the distance of our targets), whichin reality may have complex structures, so that our estimatesshould be considered as lower limits. Finally, high depletionof C18O could be due to mechanisms other than freeze-outin regions where young stellar objects are embedded, andcan decrease the CO abundance even in warm environments(for a detailed explanation, see Fuente et al. 2012). This

comment especially concerns the clumps where the averagekinetic temperature exceeds the CO sublimation tempera-ture of ∼ 20 K.

5 DISCUSSION

5.1 Virial state of the clumps

In order to investigate if our targets are gravitationallystable, we have computed the virial masses of the clumpsassuming virial equilibrium and negligible contributions ofmagnetic field and surface pressure. Assuming also that thecores are spherical, one can demonstrate that the gas massis given by:

MVIR(M⊙) ≃ k∆v2(km s−1)R(pc) , (5)

where R is the clump radius, ∆v2 the line width and k is amultiplicative factor that depends on the gas density distri-bution as a function of the distance to the clump centre (seeEq. (3) in MacLaren et al. 1988). For a homogeneous clump(i.e. ρ = const.), k ≃ 210, while for ρ ∼ r−2, k ≃ 126. Themillimetre continuum maps do not allow to derive the den-sity profile of the clumps because of the low angular resolu-tion and sensitivity to extended emission. However, previousstudies suggest that the density structure of massive clumpsreasonably can be approximated by a constant density inthe inner regions, and by a steep power-law in the outer lay-

10 F. Fontani et al.

Table 9. Parameters used to determine the CO depletion factor: integrated intensity of the C18O (3–2)line (I(C18O))), C18O total column density (N(C18O)), observed C18O abundance (XO

C18O), ’expected’

C18O abundance (XEC18O

), and CO depletion factor (fD = XEC18O

/XOC18O

). The errors are given between

parentheses. The uncertainty on fD based on the calibration errors affecting the C18O and H2 columndensities is of the order of the 40-50%.

source I(C18O) N(C18O) XOC18O

XEC18O

fD(K km s−1) (×1015cm−2) (×10−8) (×10−7)

08477–4359c1 6.5(0.1) 3.03(0.05) 2.14 1.45 713039–6108c6 1.5(0.2) 0.65(0.09) 0.95 2.08 2213560–6133c2 1.8(0.1) 1.08(0.06) 0.99 2.68 27

14183–6050c3 0.9(0.1) 0.41(0.04) 0.60 2.54 4215038–5828c1 0.89(0.09) 0.52(0.05) 0.61 3.30 5415278–5620c2 2.5(0.1) 1.10(0.04) 1.34 2.93 2215470–5419c1 2.3(0.1) 1.36(0.06) 0.99 3.46 3515470–5419c3 2.6(0.1) 1.06(0.04) 0.96 3.46 3615557–5215c2 4.2(0.1) 1.90(0.04) 1.20 3.79 3215557–5215c3 1.6(0.1) 0.76(0.05) 1.57 3.79 2416061–5048c1 8.8(0.2) 4.7(0.1) 2.83 3.39 1216061–5048c4 3.0(0.1) 1.23(0.04) 1.01 3.39 3416093–5128c2–a 3.55(0.08) 1.59(0.04) 1.72 5.09 2916093–5128c8 1.00(0.06) 0.44(0.03) 0.65 5.09 7816164–4929c3 3.01(0.07) 1.40(0.03) 0.93 2.80 3016254–4844c1 2.82(0.09) 1.93(0.06) 1.35 3.39 2516435–4515c3–a 1.73(0.09) 0.68(0.07) 0.57 4.13 7316482–4443c2 4.7(0.1) 2.96(0.06) 4.45 3.87 916573–4214c2 7.8(0.1) 2.2(0.3) 1.18 2.98 2517040–3959c1 6.32(0.09) 5.32(0.08) 3.32 1.59 517195–3811c2 3.07(0.08) 1.33(0.03) 0.75 4.04 54

Figure 3. Virial masses against gas masses computed from the

1.2 mm dust continuum emission. MVIR is calculated assuming a den-sity distribution of the type ρ = const. (left panel) and ρ ∼ r−2 (rightpanel). In both panels, the line indicates MVIR= M .

ers (see e.g. Beuther et al. 2002, Fontani et al. 2002, Hill etal. 2005).

In Fig. 3 we compare the mass derived from the dustcontinuum emission to MVIR obtained assuming ρ ∼ const.(left panel) and ρ ∼ r−2 (right panel). We note that MVIR

is on average larger than M for the case ρ ∼ const., while itis on average smaller for the other case. Because the overalldensity distribution of the clumps is likely in between thesetwo extreme cases, we suggest that, on average, the clumpsare close to virial equilibium. Our findings are in good agree-ment with those derived by Lopez-Sepulcre et al. (2010)in a sample of high-mass clumps supposed to span a widerange of evolutionary stages. The gas masses of the high-mass clumps studied by Lopez-Sepulcre et al. (2010) werederived from millimetre continuum measurements, and thedust temperatures assumed are in agreement with ours (seetheir Table 1). Moreover, they estimated the virial massesfrom C18O (2–1) line widths and assuming ρ ∼ const. Asshown in Fig. 4, we do not highlight systematic differencesbetween the two sub-samples, and between infrared-darkand infrared-bright sources either. On average, the sourcesof the Lopez-Sepulcre et al.’s sample appear to have slightlylarger virial masses, so that potentially they might be lessgravitationally bound than the clumps studied in this work.However, we stress that the parameters from which the massestimates are derived are affected by large errors (especiallythe dust mass opacity, the distance and the gas-to-dust ra-tio) and difficult to quantify. For example, the dust massopacity coefficient can introduce a factor 2 in the uncer-tainty (Beuther et al. 2002). Hence our conclusions on thevirial stability of the targets must be taken with caution.

CO depletion in infrared dark clouds 11

Figure 4. Virial masses against gas masses computed from the dustcontinuum emission for the clumps studied in this work (triangles) andthe massive clumps observed by Lopez-Sepulcre et al. (2010, squares).Filled and open triangles represent sources detected and undetectedat 24 µm, respectively, while the filled and open squares represent theinfrared-bright and infrared-dark sources selected by Lopez-Sepulcreet al. (2010), respectively. The cross indicates the clump with no 24µm image available, 13039–6108c6. MVIR is calculated assuming ho-mogeneous clumps. The line indicates MVIR= M .

5.2 24 µm dark versus 24 µm bright clumps

The presence of mid-infrared emission in molecular clumpsis very often the sign of embedded star formation activity.Therefore, the IRDCs detected at 24 µm could harbour ob-jects more evolved than those undetected. If this is the case,the two groups should have physical properties indicativeof a different evolutionary stage. In Fig. 5 we show his-tograms which compare some physical and chemical prop-erties of the IRDCs detected and undetected at 24 µm: linewidths, kinetic temperature, column (Nt(H2)), volume andsurface density of H2, gas masses (both from continuum andthe virial theorem), CO depletion factor and clump diam-eter. For MVIR, we have considered the values calculatedassuming ρ ∼ const, bearing in mind that those obtainedin the case ρ ∼ r−2 are just systematically lower by a fac-tor ∼ 1.7. For the unresolved source 16482–4443c2, we havenot included the n(H2) and Σ(H2) lower limits, and replacedNt(H2) with Np(H2). Although the statistics is poor becausethe two sub-samples contain eight 24 µm bright and twelve24 µm dark objects, the comparative histograms are usefulto check if one can notice clear systematic differences. Aninspection of the comparative histograms indicates that the24 µm dark clumps have lower Tk and higher fD. In fact,the average Tk and fD of the infrared-bright group turnsout to be ∼ 20 K and 28, respectively, while that of theinfrared-dark group is ∼ 16 K and 35, respectively. Also,if we exclude from the statistical analysis the clear outliers17040–3959c1 and 16573–4214c2, the 24 µm dark clumps

tend to have lower mass (average M of ∼ 260 M⊙ fromthe 1.2 mm continuum), smaller source-averaged H2 columndensity (average Nt(H2) of 8.9 × 1022 cm−2), and smallerH2 surface density (average Σ(H2) of 0.19 gr cm−2) thanthe infrared-bright clumps (average M of ∼ 370 M⊙, av-erage Nt(H2) of 1.3 × 1023 cm−2, average Σ(H2) of 0.27 gcm−2).

The fact that on average the group of the 24 µm-darksources has lower Tk and higher CO depletion factor than the24 µm-bright objects suggests that it likely contains objectsless evolved, and this is in accordance with other compara-tive studies of massive clumps with or without indicationsof embedded star formation (Hill et al. 2005). However, thedark sources also have lower M and H2 column and surfacedensity, and this could indicate that the 24 µm dark clumpsmay be destined to form less massive stars/clusters, thusexplaining the non-detection at mid-infrared wavelengths,although material could still be accreting.

5.3 Relation between Tk and line widths

Two indicators of active star formation are the line widthand the gas temperature, which are both expected to be-come higher with increasing star formation activity. In fact,the line widths of dense gas tracers are found to be higherin more evolved star-forming clumps than in quiescent re-gions of infrared-dark clouds (e.g. Hill et al. 2010, Raganet al. 2012), while the cores associated with protostars areon average warmer than the starless cores, both in low- andhigh-mass star forming regions (e.g. Foster et al. 2009, Em-prechtinger et al. 2009, Ragan et al. 2011). Fig. 6 showsthat there is a slight correlation between Tk and the C18Oline width. If we exclude from the statistical analysis 15038–5828c1, for which ∆V is much higher than that of the othertargets and has a spectrum with poor S/N (see Fig. B2), alinear least square fit to the data yields a slope of ∼ 0.09.Statistical correlation between Tk and line width can be in-vestigated also through non-parametric statistical methods,like the Kendall’s τ correlation coefficient. This measuresthe rank correlation, namely how two quantities are orderedsimilarly when ranked by each of them5. τ can range be-tween 1 (perfect agreement between the two rankings) and−1 (one ranking is the reverse of the other). We find τ ∼ 0.23between Tk and line width. These results suggest a faint cor-relation between the two parameters, which indicates thatthe warmer clumps tend to also have larger line widths, i.e.tend to be more turbulent. However, the large dispersioncannot allow us to draw any firm conclusion.

5.4 Relation between CO depletion and otherphysical parameters

One of the main results of this work is the high CO deple-tion factor measured in our targets, with values higher thanthose found in comparable high-mass clumps (e.g. Mietti-nen et al. 2011, Chen et al. 2011), and even in low-masspre–stellar cores (see e.g. Crapsi et al. 2005). The fact thatwe have derived fD from the C18O (3–2) transition, whichhas a critical density of 5 × 104 cm−3, certainly makes the

5 for details see, e.g., http://www.statsoft.com/textbook/nonparametric-statistics/)

12 F. Fontani et al.

2 4 6

0

1

2

3

4

5

10 15 20 25

0

1

2

3

4

5

23 24

0

1

2

3

4

5

-1.5 -1 -0.5 0 0.5 1

0

1

2

3

4

5

4 5 6 7

0

1

2

3

4

5

0 20 40 60 80

0

1

2

3

4

2 3 4

0

1

2

3

4

5

1.5 2 2.5 3 3.5

0

1

2

3

4

5

0 0.5 1 1.5

0

2

4

6

D (pc)

Figure 5. Histograms comparing some of the physical and chemical properties of the clumps detected (solid line) andundetected (dashed line) in the Spitzer-MIPS 24 µm image (see Figs. A1). In the panels showing N(H2), n(H2) and Σ(H2)we have not included the lower limits calculated for 16482–4443c2.

difference with respect to previous works in which fD wascalculated from the lower excitation transitions tracing ma-terial where freeze-out of CO is expected to be less impor-tant. In fact, freeze-out of CO and other neutrals onto dustgrains is favoured in gas characterised by low temperatures(T 6 20 K) and high densities (n > 105 cm−3), in whichthe freeze-out timescale is generally shorter than the free-fall timescale (e.g. Bergin & Tafalla 2007). In fact, high COdepletion factors (of the order of 10 or more) were measuredtowards the dense and cold nuclei of low-mass pre–stellarcores (see e.g. Crapsi et al. 2005). After the formation ofthe protostellar object(s) at the core centre, evaporation ofCO starts and the CO depletion factor drops (Caselli etal. 2002b). This theoretical prediction has been partly con-firmed by observations of both low- and high-mass densecores (e.g. Emprechtinger et al. 2009, Fontani et al. 2006).

In Fig. 7 we show fD against Tk (in the left panel) andn(H2) (in the right panel). At a first glance, the two plots donot reveal clear (anti-)correlations between the parameters.As made in Sect. 5.3, we have performed a closer inspectionof the data using the non-parametric ranking statistical testKendall’s τ (see Sect. 5.3). If we consider all points, we finda faint anti-correlation between fD and Tk (Kendall’s τ =–0.34) and also between fD and n(H2) (Kendall’s τ = –0.12). The anti-correlation between Tk and fD is consistentwith the increase of CO depletion with decreasing temper-ature, as already found in both low- and high-mass densecores (Emprechtinger et al. 2009, Fontani et al. 2006), whilethat between fD and n(H2) is the opposite of what expectedfrom chemical models. Even if we exclude from the statisti-cal analysis 16573–4214c2, which has n(H2) markedly muchhigher than that of the other members of the sample, the

CO depletion in infrared dark clouds 13

Figure 6. Gas kinetic temperature, Tk, derived from ammonia,against the C18O line widths, ∆V . The straight line is a least square fitto the data excluding 15038–5828c1, the clump with ∆V > 5 km s−1,i.e. much higher than any other clump and based on a very noisy spec-trum (see Fig. B2). The slope of the linear fit is∼ 0.09. Filled and opensymbols represent clumps detected and undetected at 24 µm, respec-tively. The cross indicates the clump with no 24 µm image available,13039–6108c6. Typical errorbars are depicted in the top-right corner.

correlation between fD and n(H2) is not observed (Kendall’sτ = –0.11). We point out, however, that all the parame-ters obtained are average values over angular regions muchlarger than the size expected for the embedded condensa-tions, while the correlations are predicted for single cores.Moreover, clumps embedded in different clouds may be af-fected by different environmental conditions, so increasingthe dispersion. In fact, Emprechtinger et al. (2009) foundthat dense cores in Perseus showed the best trends betweenthe physical parameters, whereas when including cores fromother star-forming regions the dispersion was significantlylarger.

Finally, Fig. 8 suggests that fD is not clearly (anti-)correlated to either the clump mass M and and the ra-tio MVIR/M which should show the tendency of a clumpto be stable against gravitational collapse. When excludingthe clear outlier 17040–3959c1 (whose mass is statisticallymuch larger than the rest of the sample), a slight correlationis found between fD and M (Kendall’s τ = 0.4). However,again the large dispersion of the data and the big errors onboth parameters do not allow us to claim firm conclusions.

6 CONCLUSIONS

We have undertaken a molecular line study, through obser-vations of C18O (3–2), NH3 (1,1) and (2,2), N2H

+ (3–2)and N2D

+ (4–3), of 21 IRDCs in the southern hemispherewith the aim of characterising the initial conditions of thehigh-mass star and cluster formation process. The sampletargeted was selected from the high-mass millimetre clumpsnot detected in the MSX images identified by Beltran etal. (2006), and includes sources with and without Spitzer24 µm emission. The NH3 inversion transitions have beenobserved with the ATCA. The rotational transitions of theother molecular species have been observed with the APEX

Figure 8. CO depletion factor (fD) against gas massderived from the 1.2 mm continuum (top panel) andMVIR/M mass ratio (bottom panel). The symbolshave the same meaning as in Fig. 6.

Telescope. We have detected C18O and ammonia emission inall clumps, and N2H

+ emission in all the 12 sources observedin this line. Only one source has been marginally detected inN2D

+ (4–3), which is, to our knowledge, the first detectionof this line in an IRDC. The clumps have a median mass of∼ 244 M⊙, appear to be gravitationally bound and possessmass, H2 column and surface densities consistent with beingpotentially the birthplace of high-mass stars. The most strik-ing result of the work is the high average CO depletion factor(derived from the expected C18O-to-H2 column density ra-tio compared to the observed value), which is in between 5and 78, with a mean value of 32 and a median of 29. Thesevalues, derived from the C18O (3–2) transition (which tracesgas denser than the lower-excitation lines commonly used inprevious works), are larger than the typical CO depletionfactors measured towards other IRDCs, and are comparableto or larger than the values derived in the low-mass pre–stellar cores closest to the onset of gravitational collapse.Also, a faint anti-correlation is found between fD and thegas kinetic temperature. These results suggest that the ear-

14 F. Fontani et al.

Figure 7. CO depletion factor (fD) against gas kinetic temperature (left panel) and H2 volume density (right panel). Thesymbols have the same meaning as in Fig. 6.

liest phases of the high-mass star and stellar cluster forma-tion process are characterised by CO depletions larger thanin low-mass pre–stellar cores. The clumps have an averagetemperature around 17 K. We have found marginal statisti-cal differences between the clumps detected and undetectedin the 24 µm Spitzer images, with the Spitzer-dark clumpsbeing on average colder, less massive and having lower H2

column and surface densities, but higher CO depletion fac-tors. This indicates that the Spitzer-dark clumps are eitherless evolved or destined to form stars and stellar clusters lessmassive than the Spitzer-bright ones.

ACKNOWLEDGMENTS

This publication is based on data acquired with the AtacamaPathfinder Experiment (APEX). The Atacama PathfinderExperiment is a collaboration between the Max-Planck-Institut fur Radioastronomie, the European Southern Ob-servatory, and the Onsala Space Observatory. We thank thestaff at the APEX telescope for performing the service modeobservations presented in this paper. FF is deeply gratefulto Ana Lopez-Sepulcre for providing us with the parametersof the massive cores in Lopez-Sepulcre et al. (2010).

REFERENCES

Alves, J., Lada, C. J., & Lada, E. A. 1999, ApJ, 515, 265Battersby, C., Bally, J., Jackson, J. M., et al. 2010, ApJ,721, 222

Beltran, M.T., Brand, J., Cesaroni, R., et al. 2006 A&A,447, 221

Bergin, E.A. & Tafalla, M. 2007, ARA&A, 45, 339Beuther, H., & Sridharan, T. K. 2007, ApJ, 668, 348Beuther, H., Schilke, P., Menten, K.M. et al. 2002, ApJ,566, 945

Brand, J., & Blitz, L. 1993, A&A, 275, 67Butler, M.J. & Tan, J.C. 2009, ApJ, 696, 484Busquet, G., Palau, Aina, Estalella, R., et al. 2009, A&A,506, 1183

Caselli, P. 2011, in ”Proceedings of the International As-tronomical Union”, S280, Vol. 7, pp. 19 - 32

Caselli, P., Walmsley, C.M., Zucconi, A., et al. 2002a ApJ,565, 344

Caselli, P., Stantcheva, T., Shalabiea, O., Shematovich,V.I., Herbst, E. 2002b, P&SS, 50, 1133

Chakrabarti, S & McKee, C.F. 2005, ApJ, 631, 792Chen, H.-R., Liu, S.-Y., Su, Y.-N., Wang, M.-Y. 2011,arXiv:1110.0952

Crapsi, A., Caselli, P., Walmsley, C. M., et al. 2005, ApJ,619, 379

Emprechtinger, M., Caselli, P., Volgenau, N.H., Stutzki, J.,Wiedner, M.C. 2009, A&A, 493, 89

Fontani, F., Palau, A., Caselli, P. et al. 2011, A&A, 529,L7

Fontani, F., Zhang, Q., Caselli, P., Bourke, T.L. 2009,A&A, 499, 233

Fontani, F., Caselli, P., Crapsi, A. et al. 2006, A&A, 460,709

Fontani, F., Beltran, M.T., Brand, J. et al. 2005, A&A,432, 921

Fontani, F., Cesaroni, R., Caselli, P., Olmi, L. 2002, A&A,389, 603

Foster, J.B., Rosolowsky, E.W., Kauffman, J. et al. 2009,ApJ, 696, 298

Frerking, M., Langer, L., Wilson, R. 1982, ApJ, 262, 590Fuente, A., Caselli, P., McCoey, C. 2012, arXiv:1202:5891Hernandez, A.K., Tan, J.C., Caselli, P. et al. 2011, ApJ,738, 11

Hill, T., Longmore, S.N., Pinte, C. et al. 2010, MNRAS,402, 2682

Hill, T., Burton, M.G., Minier, V. et al. 2005, MNRAS,

CO depletion in infrared dark clouds 15

363, 405Ho, P. T. P., & Townes, C. H. 1983, ARA&A, 21, 239Jimenez-Serra, I., Caselli, P., Tan, J.C. et al. 2010, MN-RAS, 406, 187

Kauffmann, J. & Pillai, T 2010, ApJ, 273, L7Krumholz, M.R., & McKee, C.F. 2008, Nature, 451, 1082Lacy, J. H., Knacke, R., Geballe, T. R., & Tokunaga, A.T. 1994, ApJ, 428, L69

Langer, W. D., Wilson, R. W., Goldsmith, P. F., & Beich-man, C. A. 1989, ApJ, 337, 355

Lopez-Sepulcre, A., Cesaroni, R., Walmsley, C.M. 2010,517, L66

MacLaren, I., Richardson, K.M. & Wolfendale, A.W. 1988,ApJ, 333, 821

Miettinen, O., Hennemann, M., Linz, H 2011, A&A, 534,134

Pagani, L., Daniel, F., Dubernet, M.-L. 2009, A&A, 494,719

Pillai, T., Kauffmann, J., Wyrowski, F., et al. 2011, A&A,530, A118

Pillai, T., Wyrowski, F., Hatchell, J., Gibb, A.G., Thomp-son, M. A. 2007, A&A, 467, 207

Pillai, T., Wyrowski, F., Carey, S.J., Menten, K.M. 2006,A&A, 450, 569

Pineda, J.E., Caselli, P., Goodman, A.A. 2005, ApJ, 679,481

Ragan, S.E., Heitsch, F., Bergin, E.A., Wilner, D. 2012,arXiv:1201.1402

Ragan, S.E., Bergin, E.A., Wilner, D. 2011,arXiv:1105.4182

Ragan, S.E., Bergin, E.A., Plume, R. et al. 2006, ApJS,166, 567

Rathborne, J.M., Simon, R., Jackson, J.M. 2007, ApJ, 662,1082

Rathborne, J. M., Jackson, J. M., Zhang, Q., and Simon,R. 2008, ApJ, 689, 1141

Rathborne, J.M., Jackson, J.M., Chambers, E.T. etal. 2010, ApJ, 715, 310

Simon, R., Rathborne, J.M., Shah, R.Y. et al. 2006, ApJ,653, 1325

Tafalla, M., Santiago-Garc´a, J., Myers, P.C. et al. 2006,A&A, 455, 577

Tafalla, M., Myers, P.C., Caselli, P., & Walmsley, C.M.2004, A&A, 416, 191

Vasyunina, T., Linz, H., Henning, Th. et al. 2009, A&A,499, 149

Wilson, T.L. & Matteucci, F. 1992, A&AR, 4, 1Wilson, T.L. & Rood, R. 1994, ARA&A, 32, 191Zhang, Q., Wang, Y., Pillai, T., and Rathborne, J. 2009,ApJ, 696, 268

APPENDIX A: SPITZER AND SIMBA IMAGES

APPENDIX B: SPECTRA

This paper has been typeset from a TEX/ LATEX file preparedby the author.

16 F. Fontani et al.

Figure A1. Superimposition of the Spitzer-MIPS 24 µm images on the SIMBA 1.2 mm continuum maps (blue contours:first contour and step correspond to the 3 σ rms level) towards the IRDCs studied in this work and observed with Spitzer-MIPS (for clump 13039–6108c6, the Spitzer-MIPS image is not available). In each frame, the red circle indicates the targetedmillimetre clump.

CO depletion in infrared dark clouds 17

Figure A1. Continued.

18 F. Fontani et al.

Figure B1. Spectra of the twelve sources observed and detected in both N2H+ (3–2) and C18O (3–2). In each frame, weshow a range of ±20 km s−1 around the systemic velocity adopted to centre the spectra (given in Table 1) and derived fromCS observations. For two objects, 16093–5128c2 and 16435–4515c3, the spectrum contains velocity components separatedby several km s−1, so that we show a broader velocity range.

CO depletion in infrared dark clouds 19

Figure B2. Spectra of the nine sources observed and detected in C18O (3–2) only. As in Fig. B1, the velocity interval onthe x-axis range from VLSR – 20 km s−1 to VLSR + 20 km s−1.